Have you seen the movie “Her”? If you are not sure we will summarize it in one sentence: a human falling in romantic love with a machine! The film shows the relationship of Theodore, a lonely writer with his AI computer operating system, Samantha.
It is considered as a science-fiction movie, but, how far is this from reality? Obviously, the idea of a romantic relationship between an AI system and a human sounds at least weird, but how was this possible?
Besides from the “is just a movie” Samantha had some features that made Thomas think she wasn’t just a machine.
The main features of artificial intelligence systems are the ability to “learn” and “problem solving”, two activities we associate with human minds and this is exactly what Samantha was able to do. But, how? How is it possible to teach a machine?
Machine Learning and Deep Learning systems have been around for a while, and they have much more applications that you may imagine, however they should pass through a learning process to develop all their potential.
The main issue is that this process is slow and is mainly based on interacting with real users to learn from those experiences. However, users get tired if after a couple of interactions the machine is not able to understand them beyond the most basic staff, this obviously is a dilemma: should we improve the machine performance decreasing user experience?
Until now, this seemed the only possible solution and not fully develp or limited AI interfaces are being put in production.
But what if we could speed up the training process of a machine? This way systems will be ready to interact with users without harming UX.
Effective training, requires effective materials, and this doesn’t mean more resources but the appropriate ones. When you go to the gym with objectives to achieve in mind, you will need to follow a specific workout routine.
For AI systems depending on what we want them to do we will need different materials:
-If we want the system to interact with users in Social Platforms such as Facebook we will need to provide them with non-formal vocabulary and text structures, so it will understand better the user’s slang.
-If we want the system to interact with users in an airline’s help center we will need to teach first the vocabulary related to aircrafts, flights etc. This reduces significantly the amount of words to teach and the time to do so. Obviously, it doesn’t take the same amount of time to learn 5.000 words than 80.000.
Summing up, creating Samantha is possible now, the technology to create an Artificial Intelligence systems with the ability to understand humans demands accurately and answer them, already exist. Probably we won’t fall in love with this system, but there is no doubt that it will become an important part of our lifes like it happened to Theodore.
A robust discussion persists within the technical and academic communities about the suitability of LLMs…
Chinese, Southeast Asian, and Arabic names require transliteration, often resulting in inconsistent spellings in Roman…
Customizing Large Language Models in 2 steps via fine-tuning is a very efficient way to…
Discover the advantages of using symbolic approaches over traditional data generation techniques in GenAI. Learn…
In the blog "General Purpose Models vs. Verticalized Enterprise GenAI," the focus is on the…
Bitext introduced the Copilot, a natural language interface that replaces static forms with a conversational,…